programatically checking failed machines

Re: programatically checking failed machines

Willy Tarreau — Wed, 06 Jun 2012 00:20:01 +0200

On Mon, Jun 04, 2012 at 01:13:31PM -0700, Cord MacLeod wrote:
> Just got around to attempting this, it appears not to work with SSL. Is there any other option (stunnel breaks a lot and doesn't support 1 off connections)?
>
> * Reloading haproxy haproxy [WARNING] 155/201028 (3442) : config : 'stats' statement ignored for proxy 'ssl-relay' as it requires HTTP mode.

That's expected since your SSL instance works in TCP mode. The "stats uri"
statement must be in an HTTP instance (use a dedicated port as most people
do, it looks like a waste at first but it's very convenient in the long run).

Also the "stats socket" in the global section allows you much more fun, as
you can connect there to get stats, change weights, start/stop servers,
kill server connections, etc...

Regards,
Willy

Re: programatically checking failed machines

Baptiste — Tue, 05 Jun 2012 09:10:01 +0200

Could you at least provide your config?

cheers

On Mon, Jun 4, 2012 at 10:13 PM, Cord MacLeod wrote:
> Just got around to attempting this, it appears not to work with SSL. Is
> there any other option (stunnel breaks a lot and doesn't support 1 off
> connections)?
>
> * Reloading haproxy haproxy
> [WARNING] 155/201028 (3442) :
> config : 'stats' statement ignored for proxy 'ssl-relay' as it requires HTTP
> mode.
>
> [ OK ]
>
>
>
> On May 30, 2012, at 12:03 PM, Miah Johnson wrote:
>
> Cord,
>
> Yes! You can use 'stats socket' or 'stats uri', and a small program that
> reads that to monitor your haproxy system. From there you can gather many
> statistics about your haproxy server, including but not limited to server
> status (up,down,etc), number of failed checks, and downtime.
>
> --
> Miah Johnson
>
> On Wednesday, May 30, 2012 at 11:36 AM, Cord MacLeod wrote:
>
> List,
>
> I'm wondering, is there a hook or something of this nature I can call to
> find if/what are the down machines in my config file? Essentially the issue
> I'm attempting to solve is to automatically remove servers that haproxy
> believes are down. These machines are all SSL, so the ssl-hello-chk is the
> only way to tell if these machines are alive. Has anyone run into this?
>
>
>

Re: programatically checking failed machines

Cord MacLeod — Mon, 04 Jun 2012 22:20:01 +0200

Just got around to attempting this, it appears not to work with SSL. Is there any other option (stunnel breaks a lot and doesn't support 1 off connections)?

* Reloading haproxy haproxy [WARNING] 155/201028 (3442) : config : 'stats' statement ignored for proxy 'ssl-relay' as it requires HTTP mode.
[ OK ]

On May 30, 2012, at 12:03 PM, Miah Johnson wrote:

> Cord,
>
> Yes! You can use 'stats socket' or 'stats uri', and a small program that reads that to monitor your haproxy system. From there you can gather many statistics about your haproxy server, including but not limited to server status (up,down,etc), number of failed checks, and downtime.
>
> --
> Miah Johnson
>
> On Wednesday, May 30, 2012 at 11:36 AM, Cord MacLeod wrote:
>
>> List,
>>
>> I'm wondering, is there a hook or something of this nature I can call to find if/what are the down machines in my config file? Essentially the issue I'm attempting to solve is to automatically remove servers that haproxy believes are down. These machines are all SSL, so the ssl-hello-chk is the only way to tell if these machines are alive. Has anyone run into this?
>

Re: programatically checking failed machines

Baptiste — Thu, 31 May 2012 06:50:01 +0200

Hi,

You could also use stunnel in client mode and run your health checks over it :)

cheers

On Wed, May 30, 2012 at 9:03 PM, Miah Johnson wrote:
> Cord,
>
> Yes! You can use 'stats socket' or 'stats uri', and a small program that
> reads that to monitor your haproxy system. From there you can gather many
> statistics about your haproxy server, including but not limited to server
> status (up,down,etc), number of failed checks, and downtime.
>
> --
> Miah Johnson
>
> On Wednesday, May 30, 2012 at 11:36 AM, Cord MacLeod wrote:
>
> List,
>
> I'm wondering, is there a hook or something of this nature I can call to
> find if/what are the down machines in my config file? Essentially the issue
> I'm attempting to solve is to automatically remove servers that haproxy
> believes are down. These machines are all SSL, so the ssl-hello-chk is the
> only way to tell if these machines are alive. Has anyone run into this?
>
>

Re: programatically checking failed machines

Miah Johnson — Wed, 30 May 2012 21:10:01 +0200

Cord,

Yes! You can use 'stats socket' or 'stats uri', and a small program that reads that to monitor your haproxy system. From there you can gather many statistics about your haproxy server, including but not limited to server status (up,down,etc), number of failed checks, and downtime.

--
Miah Johnson

On Wednesday, May 30, 2012 at 11:36 AM, Cord MacLeod wrote:

> List,
>
> I'm wondering, is there a hook or something of this nature I can call to find if/what are the down machines in my config file? Essentially the issue I'm attempting to solve is to automatically remove servers that haproxy believes are down. These machines are all SSL, so the ssl-hello-chk is the only way to tell if these machines are alive. Has anyone run into this?

programatically checking failed machines

Cord MacLeod — Wed, 30 May 2012 20:40:01 +0200

List,

I'm wondering, is there a hook or something of this nature I can call to find if/what are the down machines in my config file? Essentially the issue I'm attempting to solve is to automatically remove servers that haproxy believes are down. These machines are all SSL, so the ssl-hello-chk is the only way to tell if these machines are alive. Has anyone run into this?